Let's introduce a library to run circle-interpreter in python directly.
Why
As part of the work to make our ONE tools python friendly, running circle-interpreter in python directly is a good first step.
If a circle binary is being in the memory, it should be exported first to .circle file to run circle-interpreter because circle-interpreter only receive a circle file path.
This flow has a few problems. First, even though users just want to run circle model with numpy arrays in python interpreter, the flow forces users to dump circle file. Second, then, circle-interpreter re-loads the circle file for the inference, which causes to take double memory: the circle binary in python interpreter and the one in circle-interperter.
There is ctypes python library that allows calling c or c++ functions in DLLs or shared libraries directly. I'm going to introduce a library in circle-interpreter that can be used in python interpreter with ctypes library.
What
Let's introduce a library to run circle-interpreter in python directly.
Why
As part of the work to make our ONE tools python friendly, running circle-interpreter in python directly is a good first step.
If a circle binary is being in the memory, it should be exported first to
.circlefile to run circle-interpreter because circle-interpreter only receive a circle file path.This flow has a few problems. First, even though users just want to run circle model with numpy arrays in python interpreter, the flow forces users to dump circle file. Second, then, circle-interpreter re-loads the circle file for the inference, which causes to take double memory: the circle binary in python interpreter and the one in circle-interperter.
There is ctypes python library that allows calling c or c++ functions in DLLs or shared libraries directly. I'm going to introduce a library in circle-interpreter that can be used in python interpreter with ctypes library.